Is Mistral Large 4 available with Zero Data Retention on Elosia?

Mistral Large 4 runs on Elosia's standard backend; ZDR is available on many other models in the catalog.

How much does Mistral Large 4 cost on Elosia?

Mistral Large 4 is billed by usage at $0.6799999999999999 per 1M input tokens and $2.09 per 1M output tokens, within your Elosia plan.

What is Mistral Large 4's context window?

Mistral Large 4 supports a context window of up to 1,048,576 tokens.

Is Mistral Large 4 hosted in the EU?

Yes, Mistral Large 4 is available on Elosia's EU-hosted (EdenAI) backend for enterprise organizations.

When should you choose Mistral Large 4?

Mistral Large 4 is a flagship model, best for the most demanding reasoning, agentic and long-context tasks.

Mistral Large 4

MistralFlagship

ThinkingTool UseVisionStructured Output

About this model

Mistral Large 4 is a frontier multimodal (text and image input) model from Mistral AI built for reasoning, coding, and agentic workloads. It offers a 1M-token context window with up...

Performance Tier

Flagship

Mistral Large 4 is a flagship model from Mistral : the most capable in their lineup.

Best-in-class model from this provider. Highest performance across benchmarks, ideal for demanding tasks.

Pricing

This model is included in Elosia plans

Affordable

Low cost. Suitable for sustained use and high-volume interactions.

Type	per 1M tokens
Input (prompt)	$0.680
Output (completion)	$2.09
Cache read	$0.070

Capabilities

Context Length1.0M

Max Output Tokens262K

TokenizerMistral

Inputtext, image

Outputtext

Release DateOctober 6, 2026

Benchmarks

General Intelligence

MMLU

Not reported

GPQA Diamond

Not reported

Mathematics

MATH-500

Not reported

Programming

HumanEval

Not reported

SWE-bench Verified

Not reported

Reasoning

IFEval

Not reported

Humanity's Last Exam

35%

Multimodal

MMMU-Pro

76.4%

Where does Mistral Large 4 stand?

Compare its performance index against every other model.

View the performance leaderboard

Recommended Use Cases

CodingAnalysisResearchData ExtractionTranslation

Strengths

Hybrid instruct and reasoning model built for coding, agentic workflows and multimodal understanding, on a Mixture-of-Experts architecture with 52B active parameters out of 1.05T
Agentic coding across repository understanding and complex terminal workflows, plus business workflows that use tools and produce finished spreadsheets, slides and PDFs
Strong cybersecurity work, from reproducing and patching real vulnerabilities to analysing malware and writing detection rules
Image understanding across complex documents, charts and natural images, with visual grounding on technical drawings and large satellite images
Trained on data in more than 160 languages, including every official language of the European Union, and trained and served on Mistral's own infrastructure in Europe
512K-token context window with text and image input, function calling and structured outputs

Limitations

Served as a public preview: Mistral states the model is still being refined, and its weights are not yet released
Mistral publishes no GPQA Diamond, MMLU or SWE-bench score; the Humanity's Last Exam and MMMU-Pro scores shown come from third-party testing
Reasoning is on by default, which adds latency and tokens to simple requests